The determinants of public versus private provision of Emergency Medical Services

27 (2009)Organization 312–319 International Journal of Industrial 27 (2009) 312 – 319

Contents lists available at ScienceDirect

International Journal of Industrial Organization j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / i j i o

The determinants of public versus private provision of Emergency Medical Services☆ Guy David a,⁎, Arthur J. Chiang b a b

The Wharton School, University of Pennsylvania, United States Graduate School of Business, Stanford University, United States

a r t i c l e

i n f o

Article history: Received 20 January 2008 Received in revised form 24 September 2008 Accepted 6 October 2008 Available online 11 October 2008 JEL classification: L33 L24 H44 I11 Keywords: Public and private enterprises Contracting out Emergency Medical Services

a b s t r a c t Competition for the provision of local public services often involves mixing private firms and public agencies. Predicting competitive outcomes therefore requires identifying the productive and strategic advantages of different organizational configurations: pure public, pure private or a public–private mix. We consider a make-versus-buy decision in a government procurement context by identifying the strength of public agencies as having an inherent advantage in accessing local infrastructure while private firms are identified as having a superior incentive to exploit returns to scale technologies due to their ability to service multiple localities. We focus on the choice of system configuration for the provision of Emergency Medical Services (EMS), a socially important service which benefits from infrastructural synergies as well as technological improvements (i.e. medical quality). We test our predictions on a panel data set of the 200 largest US cities and find that smaller cities and poorer access to hospitals favor the mixed public–private configuration in the provision of EMS. © 2008 Elsevier B.V. All rights reserved.

1. Introduction To ensure the provision of a specific service, local governments can often choose between contracting with a private firm or providing the service through a municipal agency. The factors which determine the local government's choice of provider type have been studied from several perspectives. One approach assumes that differing residual control rights of managers operating under public and private ownership will lead the managers to behave differently under incomplete contracts (e.g. Williamson, 1985; Hart et al., 1997). Whether the public or private behavior is preferable depends on characteristics of the particular service. Levin and Tadelis (2007), develop a framework in which local governments choose provider type based on the relative importance of contracting/monitoring complexity and the value associated with quality of the service. Brown and Potoski (2003) demonstrate that high levels of risk of contract failure lead local governments to engage in a variety of monitoring techniques to improve their ability to monitor and correct vendor performance. However, private firms that contract with multiple municipalities develop a reputation that may mitigate the value of scrutiny. ☆ We are grateful to William Baxt, Tanguy Brachet, Charlie Branas, Dennis Carlton, Ed Dickinson, Reena Duseja, Scott Harrington, Thomas Hubbard, Crawford Mechem, Mark Pauly, and Steve Tadelis for helpful comments. We benefited from comments by workshop participants at the University of Illinois at Chicago, Tel-Aviv University and participants at the 2006 meeting of the American Society of Health Economists. Phil Saynisch provided exceptional research assistance. Financial support from the Leonard Davis Institute for Health Economics is gratefully acknowledged. ⁎ Corresponding author. E-mail address: [email protected] (G. David). 0167-7187/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.ijindorg.2008.10.001

In this paper, we suggest an alternative make versus buy theory of the determinants of public versus private provision in a government procurement context. We consider city-level provision of services either through in-house provision by a public agency, contracting out with a private firm, or a mixture of the two. Miranda and Lerner (1995) refer to the later case as benchmarking, in which a local government contracts out a portion of the service while producing the reminder through in-house production. While our conceptual framework is robust to the inclusion of contract incompleteness, our differentiation between public and private providers is not dependent upon it. Using this approach, we do not make any assumptions about a priori differences in behavior and objectives between public and private providers. Nevertheless, we identify a dimension along which public and private entities differ fundamentally: private firms are free to provide services to multiple cities and communities (Donahue, 1989), while a public agency is restricted to its particular city of operation.1 This in turn generates differences between public and private providers in the size of population they serve, which suggests a discrepancy between the two provider types in their access to economies of scale by aggregating service delivery over a range of cities. As the private firm can serve a larger population, it can reduce its average cost of capital, technological research, and other scaleinvariant investments, in turn raising the optimal level of investment

1 While there some examples of intergovernmental contracting across municipalities there is little cooperation among geographically disperse cities in service delivery (Warner and Hefetz, 2003).

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319

in capital and innovation development for private firms. The attractiveness of these economies of scale to a city will depend in large part on the extent to which frequent implementation of technologies and innovations are vital to the specific public service. Moreover, to the extent that scale advantages are local, the proximity of other cities served by the private firm will also affect the cities' choices of provider type. In addition, we link the optimal provider type to city level characteristics, such as population, local infrastructure, and the size and proximity of neighboring communities. Because such characteristics vary across cities, our theory can support the existence of a mixed equilibrium in which some cities may choose to provide a service through a public agency, some through a private firm, and some through a mixture of public and private providers. We develop our theory in the context of a vital and largely understudied public service, Emergency Medical Services (EMS). EMS refers to the pre-hospital medical care administered at the scene of a medical emergency and en route to a medical facility as well as to the transportation service itself. It is the “first line of defense” against death and disability for many individuals who sustain traumatic medical events, including cardiac arrest, stroke, and major wounds associated with motor vehicle crashes and violence. In addition to its crucial role in public health, EMS provides a rich environment to investigate the determinants of public versus private provision for complex services. Both public and private providers are represented and there exists variation of provider type across localities and over time. For example, in 2005, approximately half of the largest 200 U.S. cities relied solely on public provision of EMS, while the other half provided EMS by contracting out a portion of the service while producing the reminder through in-house production (JEMS, 2005). Contracting out all elements of EMS to a private EMS firm is rare among the largest 200 U.S. cities. Additionally, some cities changed their EMS providers over time; the 1990s witnessed a significant increase in the percentage of cities incorporating private providers. This trend coincided with widespread consolidation among local private providers, which resulted in the creation of a few nationwide firms that now serve a large number of urban and rural areas.2 This private consolidation movement appears consistent with our conceptual framework, since it is likely the result of changes in EMS technology and medical standards giving rise to strong scale advantages. In its 1997 Alternative Service Delivery Survey, the International City/county Managers Association (ICMA) found about 40% of U.S. cities to report using a non-government based ambulance service.3 To learn about the public/private decision from the EMS industry, it is important to understand the fundamental features of the service. One of the most important and unique attributes of EMS is the twotiered structure utilized in most cities, consisting of first response and transport. In EMS care, a first-responder arrives on the scene of an incident quickly, followed later by a paramedic-equipped ambulance, which has the responsibility of transporting the patient to a suitable hospital. The responsibility of the first responder is to arrive at the scene as soon as possible to assess the situation and provide basic, stabilizing medical care. It is especially imperative that the first responder treat problems for which a patient's condition deteriorates rapidly with time, such as cardiac arrest. The personnel providing first response is often trained only in basic life support (BLS), such as cardiopulmonary resuscitation (CPR) and the use of defibrillators. The transport ambulance, on the other hand, is usually staffed by a highly trained paramedic and equipped with advanced life support (ALS) technology, so that during transport more sophisticated medical procedures can be performed. The provider entity chosen to provide first response in a given city can be different from the provider selected to transport the patient to the hospital. An example of public–

private mix is when a public agency such as the city fire department provides first response services while a private EMS firm provides transport service. A pure public model refers to the case in which a public agency operates in both tiers. Because the primary role of first response is to arrive rapidly enough to prevent death and serious health damage resulting from the most time-critical conditions, there is an overwhelming premium for speed in first response. Fire departments already invest in the infrastructure necessary to provide rapid response to structural fires, which have a similar geographical distribution of occurrence to EMS events, and for which rapid response time is crucial. Therefore, it is logical that the vast majority of municipalities use fire departments for first response.4 In transport EMS, relatively complex medical procedures are often performed while the patient is en route to a hospital. Thus, quality of service, defined as highly trained personnel, advanced medical technology, and system management, take on larger importance in transport-tier EMS than in first response. When access to hospitals worsens (e.g. due to hospital closure) in a city, the quality of transport medical care becomes even more critical as this increases average transport time and hence the amount of time a patient will be treated by the transport EMS team before reaching the hospital. Because of the ability to serve multiple cities and hence to enjoy reduced average costs of capital and technology due to scale, private firms generally have an advantage in providing a higher level of quality improvements than public agencies. While quality is important for transport, there still exists a quality-time tradeoff; the sooner the transport ambulance arrives at the scene of an EMS incident, the sooner the patient will arrive at a hospital, where the highest level of medical care can be provided. While private firms have a relative advantage in implementing operational upgrades, public agencies have relative advantages in accessing existing infrastructure, which determines the time from call to arrival at the hospital.5 Moreover, the choice of EMS delivery modality given the tradeoff between infrastructure and scale advantages depends on characteristics of the individual city. We argue that a smaller population, reduced access to hospitals, and size and proximity of neighboring cities will increase the city's propensity to privatize the transport portion of its emergency medical services.6 The paper is organized as follows: Section II develops our conceptual framework of EMS provision in cities which centers on the interaction of scale and infrastructure with key city characteristics. Using this framework, we narrow the set of potentially optimal configurations for EMS and obtain testable predictions on the effect of city parameters on the choice of EMS provider configuration. In Section III we introduce and describe a panel of the 200 largest cities in the United States in 1991, 1998, and 2005, which includes information on each city's first response and transport provider, as well as demographic characteristics, hospital infrastructure, unionization and crime statistics. In Section IV, we test the implications about the determinants of EMS system design that arise in our framework by estimating several discrete choice models. Section V concludes the paper and discusses potential generalizations of our framework beyond EMS. 2. A theory of the determinants of EMS provision in cities In this section we develop (informally) a theory of how a local government determines its choice of providers for first response and 4

In 2005, 96% of the 200 largest US cities used fire departments for first response. Fire departments generally face a lower cost of transport infrastructure because they are already providing first response and hence there exist synergies arising from continuity of provider type, which reduce transport preparedness costs and streamline service. 6 The potential effects of additional city characteristics on the choice of EMS provider type are discussed in detail in Chiang et al. (2006). 5

2 The three major national private EMS firms are American Medical Response (AMR), Rural/Metro Ambulance Service, and Southwest Ambulance Service. 3 ICMA is an organization dedicated to fostering inter-municipality cooperation by offering various forms of consulting with city and county managers.

313

314

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319

transport Emergency Medical Services (EMS).7 We begin by assuming that the role of first response and/or transport provider can be performed by one of two players, the fire department (a public organization) or a private firm. We will refer to a city's provider choice by an ordered pair, for example, the combination of the fire department as first responder and a private firm as transport provider is denoted “public–private”. In principal, at both tiers of EMS (first response and transport) the two dimensions that are crucial for a patients' health are quality of care and speed (which is determined by the provider's level of infrastructure). And it is along these two dimensions that public and private EMS providers have fundamental competing advantages. We assume that the local government ultimately chooses the provider type(s) which result in levels of infrastructure and quality that yield the highest benefit to its community.8 Next, we discuss the particulars of the first response and transport tiers of EMS, the relative advantages that public versus private providers have in each tier, and how these factors interact with city characteristics to determine the equilibrium provider choices for a given city. The first response provider responds to all EMS calls in the city and provides basic, stabilizing medical care. Reducing the time to arrival, or response time, for first responders is vital because the health condition of an individual who sustains an incident of trauma, such as a heart attack, stroke, or wound from physical violence or a motor vehicle crash, often deteriorates rapidly with time. Conversely, more sophisticated medical capabilities are unlikely to be nearly as crucial as response time at this stage, for two reasons. First, in the majority of time-critical cases, heart attacks and heavy trauma, basic procedures (especially CPR and defibrillation) are indeed the best first step (Persse, 2002). And second, since the transport EMS provider is likely to provide more advanced care, any extra medical capabilities of first responders would only be potentially utilized in the short time window between the arrival of first response and the arrival of transport. Thus, in first response, the choice between a public or private provider largely comes down to who can achieve a faster average response. Now, in a vast majority of cities, fire departments (which are virtually always public entities) have long had a well-established fire station network, with vehicles and personnel already on-hand 24 h a day to respond to fires (Eckstein and Pratt, 2002); thus, fire departments have an existing infrastructure, both physical and human, that can be relatively easily enhanced so as to enable them to become EMS first responders in addition to firefighters. Conversely, a private firm has no such inherent infrastructure in a given city. Therefore, the costs associated with adding first response EMS to existing fire department infrastructure are likely to be (substantially) lower than creating a private provider-based first response EMS infrastructure. For this reason, we expect to observe that the vast majority of cities will employ their public fire department to provide first response EMS. That is, of the four possible provider combinations, we expect that pure-public and/or mixed public–private pairs will predominate. When EMS transport arrives at the place of emergency, the first responders “hand-off” the patient to the transport team, whose primary purpose is to bring the patient to a trauma center hospital, where the highest level of medical care is available. Since a faster arrival time by the transport team necessarily means the patient will arrive at the hospital sooner, a provider's potential level of infrastructure is again crucial for the transport tier of EMS, just as it was for first response. However, because at this stage basic procedures have already been performed by first responders and because transport time to the hospital may be

7 A formal model of this decision, which draws upon spatial economic models to explicitly capture the competing effects of infrastructure and quality improvements, is available from the authors. 8 We are assuming alignment of objectives between the government and the voting population it serves. We briefly explore the possibility of distortions from this due to the political influence of public labor unions in the empirical section.

relatively long, the quality of advanced medical procedures on-scene and en route to the hospital is potentially very important.9 Therefore, to the extent that the quality of medical care provided by the transport provider is a function of technological investments (defined broadly, as we will below), the provider that is more efficient in developing and/or implementing such technology has an advantage along the quality dimension. If the transport provider is the same as the first response provider, we assume that transport infrastructure will be less expensive (as compared to the case where the transport provider is a different entity than the first response provider, e.g. a mixed public–private system) due to operational synergies, including utilizing the same physical stations for first response and transport, personnel providing both first response and transport medical support, superior communications between first response and transport, etc. Since we expect only pure-public and mixed public–private pairs to prevail, in practice, only the public fire departments will have the possibility of enjoying these synergies. Thus, the infrastructural advantages of fire departments in first response translate into similar infrastructural, and hence response time, advantages in transport provision. The technology that improves the quality of medical care administered by transport paramedics may take several different forms. Examples include the development of ambulance system management technology, such as Automatic Vehicle Locators (AVL) which improve transport team coordination and expedite travel to the hospital, the adoption of cutting-edge medical devices in ambulances and/or purchasing new ambulances to replace older vehicles, and providing advanced life support paramedic training. What is common to many of these quality improvements is that the average cost of investment diminishes with the scale of service of the provider; that is, there are economies of scale for technological investment. By scale we mean the size and/or geographic proximity of the population a given provider services. Larger providers have more bargaining power with vendors and hence face lower costs when purchasing new vehicles, equipment, and supplies. Similarly, the cost of research developing innovations or choosing which new technologies to implement is scale-invariant, yet the returns to the developments of research increase with the number of people who will benefit from their use. Finally, training personnel in advanced procedures can be done at a location central to several municipalities, thus reducing redundant teaching and facilities costs. For a fire department, its scale is simply the population of the city it services; in practice this is essentially the limit on the scale of public, city-based agency services.10 Conversely, a private firm can service multiple cities, so their scale of service is not limited by the population of any given city. Rather, the scale of service for a private provider, as it pertains to a given city, can roughly be thought of as the weighted sum of the populations of all cities the firm serves, with weights being higher on the populations of cities that are closer to the city in question. Therefore, for a given city, a private transport provider, serving multiple cities, will be able to offer higher quality transport medical care than that of its public fire department, due to economies of scale. More precisely, the greater scale of service of a private firm will translate into lower costs for any given level of quality investments, as compared to a public provider. 9 Average transport time for a city is roughly proportional to the number of trauma center hospitals accessible to that city. Since this will vary substantially across cities, transport time will serve as a key comparative static in our theory and, subsequently, in our empirical analysis of EMS provider choice. 10 Municipalities who use in-house (i.e. public) provision of services can and sometimes do benefit from economies of scale by cooperating and sharing services with other cities. The types of intergovernmental shared services range from sharing information to establishing cross city joint organizations with shared budget and management oversight. However, with the exception of disaster planning, collaborative efforts in EMS are limited to short term informal relationships, in which each city functions separately with no commonly defined mission, structure, or effort. In some cases intergovernmental contracts in EMS between small cities and their county do exist (e.g. city of Olympia, WA and Thurston County) but never between geographically dispersed cities, as was the case for the private firms we studied.

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319

Then under a range of contractual situations, including: observable costs, contractible quality investments, and incomplete contracts with unobservable costs (but publicly known benefits and identical division of surplus for public and private providers) these cost advantages will allow a private firm to offer a more attractive ex ante contract for quality improvements than the public provider. However, the extent of this quality differential will depend on the size of the city in question relative to the (proximity-weighted) size of the outside population that the competing private firm serves. For example, consider a city with a relatively small population that is in close proximity to several other cities of comparable size; the fire department of this city will have relatively weak economies of scale, however, a private firm that services many of the cities in the area will have significantly stronger scale advantages. Conversely, consider a highly populous city that has no substantial neighboring cities; then a private firm's proximity-weighted scale advantage as compared to the city's fire department may be negligible. Also, since the purpose of advanced medical care in transport EMS is to reduce the health costs of time that occur while bringing the patient to a hospital, the value of the superior quality level offered by private firms is greater than the fewer trauma center hospitals present in a city, as this increases average time in transport. In light of the framework for infrastructure and quality investments presented here, we can summarize our predictions for EMS provider choices. Due to the premium for rapid speed in first response and the infrastructural advantages inherent to public fire departments, we expect to observe essentially only the pure-public and/or the mixed public–private provider combinations, as noted earlier. However, which of these two will ultimately obtain for a given city will hinge upon which competing advantage, synergies and infrastructure for the pure public combination or superior quality investments of the private transport provider, is stronger for that city. This will relate to a number of city characteristics. For example, in more populous cities the public provider can more closely approach the economies of scale in quality investments of a private firm, so we expect more larger cities to have a pure-public provider, while smaller cities will prefer mixed public–private configurations. Conversely, the more significant neighboring cities a given community has, the more likely a private firm will be able to replicate its service and enhance its economies of scale advantages, thus giving rise to more public–private combinations. Finally, cities with more hospitals have shorter average EMS transport times, so the value of quality investments is relatively lower, so such cities more likely to utilize pure-public configurations. 3. Data and summary statistics Following the discussion of the framework above, to estimate the effect of city characteristics on the choice of EMS provider, we use citylevel data on the type of EMS contractors (public or private), hospital infrastructure, as well as demographic and geographic data. This section describes the three main sources of data used in this study: the Journal of Emergency Medical Services (JEMS) annual 200-city survey, the American Hospital Association (AHA) annual survey, and the US Census. We then discuss our classification of EMS contracts, describe the derivation of variables used in estimation, and provide summary statistics. 3.1. General data description The Journal of Emergency Medical Services (JEMS) annual 200-city survey is the primary component of the data. The data used here are taken from the 1991, 1998 and 2005 annual surveys.11 The cities surveyed in each of these years are the 200 most populous U.S. cities in that year, as defined on the basis of population by the U.S. Census 11 Prior to the 1991 JEMS annual survey only the 100 most populous cities were represented.

315

Bureau. The transience of the U.S. population has resulted in some cities being added and others being dropped across the three surveys. 170 cities appear in the largest 200 cities in all three survey years, and we refer to these cities as the “Full Sample.” We will also focus on the 1991–1998 time frame (which we refer to as the “First Period”) and the 1998–2005 time frame (i.e. the “Second Period”) we have a panel of 184 cities in each of these samples. For each year in our panel, we augment the EMS provider information with the following data: city and Metropolitan Statistical Area (MSA) population, land area, and income per capita, from the U.S. Census Bureau; hospital data at the city level from the American Hospital Association (AHA); aggravated assault and murder rates from the FBI crime database; and MSA-level public employee unionization figures from the Bureau of National Affairs' Union Membership and Earnings Data Book (Hirsh and Macpherson, 2003).12 City population is a key determinant of provider-type choice in our framework. While we focus on the 200 largest cities in the U.S., city population varies from less than 95 thousands to more than 8 million in our sample.13 The effect of scale on the propensity to use a public–private model draws on MSA population to create a measure of the scale advantage of a private firm in the proximity of a city. Land area is included to proxy for the average distance traveled by EMS providers and income per capita controls is for the possible influence of prohibitive expense of either public or private providers. Data on a city's hospitals are important in estimating the level of access that EMS patients will have to sophisticated care. Our crime data controls for the possibility that high levels of violent crime could deter a private firm because of excess costs that a dangerous city imposes. Omitting such variables will bias the estimates, if for example; bigger cities also have more crime. Alternatively, crime leads to injuries that increase the demand for EMS in the city independent of population size. Finally, the inclusion of unionization data serves as a control for the potential effect that cities with high levels of public service unionization would tend towards public provision, as city officials, who ultimately choose the provider type, want to receive political support from these unions (for more detailed discussion see Hart et al., 1997). Fernandez (2005) finds that the strength of public unions and size of the public work force influence the likelihood of contracting out. 3.2. Classification of EMS contracts The JEMS survey is the leading source of EMS provider type information for U.S. cities. Table 1 summarizes the different firstresponse and transport providers' combinations in the full, first and second samples.14 Table 1 shows that fire departments have been and remain virtually the only provider of first response in U.S. cities. In transport service, on the other hand, there is a diversity of provider types. This is consistent with our theoretical prediction that the salient first response/transport provider combination used will be either the pure public or the public– private ones. We observe a strong trend towards privatization of transport EMS between 1991 and 1998, and to a lesser extent between 1998 and 2005. Table 2 exhibits EMS provider configuration transition matrices between 1991 and 1998 and between 1998 and 2005. Between 1991 and 1998, 17 cities switched from a public based transport provider to a private one while only two switched from a public–private model to a pure public one; between 1998 and 2005 nine additional cities switched from a pure public to a public–private combination, however, six cities exercised “contracting back in,” that is, returned to a pure public

12 If data for a variable for a particular year was not available, we used data from the closet possible year for which it was available. 13 Even though there is great variability within the 200 largest cities, one might expect the scale-infrastructure tradeoff to be more pronounced in smaller cities. 14 We treat local quasi-public models for transport EMS, such as public utility model as public.

316

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319

Table 1 Count of cities by type of first response provider and transport provider for 1991, 1998, and 2005 in each of the three samples EMS combination

Full sample

First sample

Second sample

1991

1998

2005

1991

1998

1998

2005

Public, public Public, private Private, private Private, public Total

120 50 0 0 170

105 65 0 0 170

102 68 0 0 170

125 59 0 0 184

110 74 0 0 184

110 74 0 0 184

107 77 0 0 184

Table 2 Transition matrices for cities between 1991 and 1998 and between 1998 and 2005

provision of services model (Hefetz and Warner, 2004). In addition, 10 out of the 16 cities entering the top 200 populous cities list between 1991 and 1998 used a mixed public–private EMS configuration compared with 7 out of 16 cities entering the top 200 populous cities list between 1998 and 2005. While no city in our sample relies solely on volunteer organizations, five cities augmented their EMS force with volunteer organizations.15 In addition, both pure-public and public–private systems rely on volunteer manpower (Chiang et al., 2006).16 Volunteers may differ from paid personnel on various dimensions and may favor one system configuration over another. Our data does not allow us to address this issue. 3.3. Construction of scale and access measures In order to test the predictions outlined in Section II we need empirical measures of hospital access and population of a city's relevant neighborhood, as these are important parameters influencing the relative importance and strength, respectively, of a private provider's scale advantages. Our construction of such measures is described below. 3.3.1. Proxy for hospital access To proxy access to hospital we use data from the American Hospital Association (AHA) annual surveys. For each city i and year t, we calculate the number of general community hospitals with 50 or more licensed beds and indicating use of emergency room.17 We then divide the number of hospitals in each city, i, and year, t, by its land area in year t, as reported by the U.S. Census Bureau. Hospitals per Sq Milei;t =

# Hospitalsi;t Land Areai;t

This measure fits well conceptually with our theory, which predicts that the importance of innovations (associated with private providers) falls with an increase in the number of hospitals, keeping land area constant. This measure assumes that, conditional on land size and the number of hospitals, their geographical spread is similar across cities.18

15 The five cities include: Columbus, OH, Houston, TX, Flint, MI, Rochester, NY, and Virginia Beach, VA. 16 The ICMA survey indicates 10% of respondents mentioned some reliance on volunteers. 17 Trauma designation levels may be a more natural way to assess the city's access to acute emergency care. We find the AHA data on trauma hospitals' designation levels in 2004 comparable with data from the Trauma Information and Exchange Program (TIEP). TIEP data does not exist for the years 1991 and 1998. 18 To insure that this simplifying assumption is reasonable we used ARCView 9.2, a GIS mapping software, to construct a more precise access measure. For each hospital in a city we draw a circle around it with a 5 miles radius, then summed up the nonoverlapping area under the circles and divided it by the city's land area. We constructed this measure for a subsample of 24 cities. We find that while this computation does not rely on assumptions about geographical spread of hospitals it produced similar measures.

3.3.2. Deriving a measure of private transport economies of scale As discussed in Section II, scale advantages from private EMS provision, faced by a city is an equilibrium notion. Focusing on local economies of scale in transport, such as centralized training, maintenance and support, we track the population of the metropolitan statistical area for each city. In an isolated city, the fact that a public transport provider is limited by the city's boundaries will have a smaller effect, as the potential scale advantages from contracting with a large private firm are limited by the dearth of neighboring cities. In practice, there are two ways to derive a measure of (local) scale advantages: the first is based on the idea that cities assume that their neighboring cities will maintain their EMS provider type, hence creating a link between the city's current choice and the lagged choice of its neighboring cities. Under this approach, privatization by some cities in period t − 1 may lead neighboring cities that did not privatize their transport services in t − 1 to do so in period t. Essentially, this means that cities are not choosing their EMS provider strategically (i.e. firms are not incorporating their neighboring cities best-response functions in their optimization problem). The second approach, to which we alluded in Section II, builds on the assumption that cities choose their EMS provider incorporating the best-response of their neighboring cities. Each city can calculate the equilibrium expected aggregated population served by a private multi-city provider in its area. The greater this expected population is the larger the economies of scale brought by a private EMS transport provider. Note that on the one hand, cities making a smaller fraction of their metropolitan area face higher potential benefits from privatization; however, the larger the population share of the MSA a neighboring city makes up, the lower its propensity to privatize. Thus the effect of a larger “residual” MSA population (i.e. the fraction of an MSA population not living in the city of interest) on a city's decision to privatize is dependent upon the distribution of the residual population across the other cities in the MSA. To see this, assume two equal size metropolitan areas. The first metropolitan area encompasses two cities with population shares 20% and 80%, while the second metropolitan area encompasses three cities with population shares 20%, 40%, 40%. Let sim denotes the size of city i in MSA m. Note that s11 = s22, and s21 = s22 + s32. The probability of utilizing a public–private model, p(sim), falls with population due to the scale arguments made earlier, so that: p(s11) = p(s12), and p(s22) = p(s32) N p(s21). When considering the two small (20%) cities, s11 and s12, the expected local population

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319 Table 3 Mean comparisons between cities using public–private and cities using pure public EMS configurations by year 1991 Public– private Observations Population

65 182,062 (152918) Share of MSA 0.250 (0.219) Herfindahl Index 1555 (1337) Hospital access (per 76.88 1000 square miles) (52.94) Murder rate 10.07 (8.70) Aggravated 626.27 assault rate (388.3) Household income 30,936 (8268) Unionization rate 41.25 in MSA (17.99)

1998

2005

Pure public

Public– private

Pure public

Public– private

Pure public

135 382,861 (746041) 0.352 (0.239) 1759 (1807) 91.23 (86.95) 17.23 (14.78) 721.16 (482.5) 27,751 (7386) 34.58 (18.18)

84 182,801 (114991) 0.209 (0.197) 1179 (1066) 59.14 (48.70) 8.25 (6.17) 499.73 (360.9) 42,696 (12,060) 42.35 (16.49)

116 433,629 (793694) 0.267 (0.207) 1394 (1358) 80.20 (73.82) 12.85 (11.85) 569.83 (366.1) 38,372 (9649) 35.82 (17.48)

84 219,251 (168209) 0.221 (0.198) 1313 (1087) 49.68 (38.40) 8.80 (8.28) 389.69 (247.2) 47,681 (14,693) 43.74 (18.97)

116 474,163 (872272) 0.265 (0.227) 1518 (1561) 68.14 (62.83) 10.41 (9.84) 433.87 (272.2) 43,206 (11,757) 34.48 (17.86)

served by a private provider, Em is greater in the three cities MSA,19 so the 20% city in this MSA is more likely than the 20% city in the two-city MSA to combine public and private EMS, due to higher expected scale advantages from incorporating a private transport provider. Now assume that the first metropolitan area encompasses two cities with population shares 10% and 90% (instead of 20% and 80%). Note that s11 + s21 = s12 + s22 + s32, s11 b s12, and s21 N s22 + s32, and hence p(s11) N p(s12) N p(s22) =p(s32) N p(s21). Note that in this case we can no longer conjecture which city will be more likely to choose a public–private EMS combination in equilibrium. The smallest city in the first MSA faces a larger residual market (90%) yet a more concentrated one (a single competitor). It is possible for cities to achieve economies of scale through either cooperation or benchmarking. Therefore, geographical proximity of cities creates incentives for inter-municipal cooperation as well as contracting out portions of the service (Sclar, 2000). This is especially true for cities that make a smaller share of their metropolitan area's population, and may explain why suburbs use higher levels of contracting compared with adjacent large cities (Hirsch, 1995). This calls for controlling for geographical proximity. To capture the effect of the size of the residual market on the city's privatization decision we calculate for each year the share of each city's population out of its MSA population. To capture the effect of concentration on the city's transport privatization decision we calculate an MSA-level HHI based on the population distribution across its major cities. We use these two offsetting measures as well as their interaction to elicit the expected payoff from employing a public–private combination due to the potentially greater scale advantages of a private firm in its city's MSA. 3.4. Summary statistics: city characteristics There are significant differences in characteristics between cities that employ private–public EMS systems and those employing pure public ones. Table 3 presents mean comparisons between those cities for each of the sample years: 1991, 1998 and 2005 and for all 200 cities. Cities using a pure public EMS system are significantly larger, have higher population share of their metropolitan statistical area population, have more hospitals per 1000 square mile, have higher murder and aggravated assault rates, lower median household income and public service unionization and more concentrated MSAs, as com19 p(s22) N p(s21) ⇒ p(s22)· s21 N p(s21)· s21 ⇒2· p(s22)· 0.5· s21 N p(s21)· s21 ⇒ 2· p(s22)· s22 N p(s21)· s21 and since s22 = s32 and p(s22) = p(s32) ⇒p(s22)· s22 + p(s32)· s32 N p(s21)· s21 □

317

pared to cities which employ a public–private EMS system. These differences between public and private utilizing cities are consistent with both our and other existing theories. The next table exhibits the time trends for the key city level variables for the full, first and second samples. From Table 4, and particularly from the Full Sample, it is clear that many of our variables of interest exhibit significant trends over time. As noted earlier, the share of cities integrating private EMS transport providers into their system increased by 30% from 1991 to 1998 and more gradually (by 5%) from 1998 to 2005. In terms of city characteristics, population size increased during the sample, while the number of hospitals per square mile, murder and aggravated assault rates dropped considerably from 1991 to 2005. The average for cities' population share and concentration of their MSA, as well as public service unionization at the MSA level fluctuates somewhat yet remains relatively steady. 4. Estimation and results Our binary dependent variable equals 1 when a city uses a public– private EMS provider mix and 0 when it uses a pure public model. We estimate the model for the unrestricted 200 cities repeated crosssections, the full 170 cities panel, and the first and second 184 cities panels. While we are interested in accounting for systematic differences across cities, a fixed effects probit analysis may lead to an incidental parameters problem. Therefore, we treat the individual city effects as unobservable random variables. To consistently estimate the model parmeters we assume that the variables and individual effects are independent and that the latter has a normal distribution. This allows us to estimate a random effects probit model and calculate average partial effects. Table 5 summarizes the results from this regression on the four explanatory variables of interest. In addition, results from a crosssectional OLS regressions and city fixed effects regression are reported. Standard errors are clustered at the municipality level to allow for correlation in EMS configuration over time within the same city. For robustness we present the results from cross-section and fixed effects regressions as well. Supporting our theoretical predictions, we find a negative effect of population size, hospital density, and the interaction between the city share of MSA and population concentration in the MSA (HHI) on the likelihood of choosing a mix public–private EMS configuration. These effects are, for the most part, economically and statistically significant. These results are robust to using spline regressions for both the city population share of its MSA and for the MSA-level HHI. Not surprisingly,

Table 4 Summary statistics for city characteristics by sample and year Full sample 1991 Share of cities w public–private Population

0.294 (0.457) 355.0 (671.7) City's share of 0.281 MSA pop. (0.226) Herfindahl Index 1640 (1538) No of hospitals per 84.53 1000 square mile (72.19) Murder rate 15.3 per 100,000 (13.1) Aggravated 700.5 assault rate (473.7) Unionization rate 36.67 in MSA (18.19) Obs. 170

First sample

Second sample

1998

2005

1991

1998

1998

2005

0.382 (0.487) 365.9 (666.2) 0.255 (0.205) 1352 (1287) 72.62 (66.03) 11.0 (9.6) 558.4 (380.9) 38.02 (17.36) 170

0.400 (0.491) 405.6 (735.1) 0.270 (0.219) 1527 (1459) 65.57 (56.72) 10.6 (9.5) 443.8 (266.2) 38.22 (19.01) 170

0.321 (0.468) 336.6 (648.7) 0.284 (0.224) 1642 (1523) 85.33 (71.45) 15.3 (13.5) 690.2 (461.6) 36.72 (18.29) 184

0.402 (0.492) 346.4 (643.7) 0.258 (0.204) 1358 (1277) 73.31 (65.53) 11.2 (10.3) 550.8 (370.4) 38.09 (17.46) 184

0.402 (0.492) 347.2 (643.5) 0.241 (0.205) 1302 (1256) 69.97 (64.94) 10.7 (9.4) 546.1 (375.9) 38.65 (17.14) 184

0.418 (0.494) 386.4 (709.6) 0.255 (0.218) 1475 (1423) 63.19 (55.87) 10.2 (9.4) 432.0 (263.8) 39.08 (18.97) 184

318

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319

Table 5 Results from OLS, random effects probit and fixed effects regressions of pure-public/ public–private combination choice Unrestricted sample

Full sample

1991–1998 sample

1998–2005 sample

−0.2005 (0.0291)⁎⁎⁎ −0.0978 (0.0341)⁎⁎⁎ −0.0639 (0.0293)⁎⁎ −0.378 (0.1798)⁎⁎

−0.185 (0.0306)⁎⁎⁎ −0.1294 (0.0384)⁎⁎⁎ −0.0614 (0.0316)⁎ −0.2785 (0.1921)

−0.1731 (0.0365)⁎⁎⁎ −0.0856 (0.0438)⁎ −0.0509 (0.0385) −0.1715 (0.2298)

−0.224 (0.0367)⁎⁎⁎ −0.1633 (0.0477)⁎⁎⁎ −0.0592 (0.0406) −0.3351 (0.2308)

Random effects probit model Log population −1.9641 (0.4238)⁎⁎⁎ Hospitals per −0.8332 1000 square mile (0.4352)⁎ City pop share of MSA −0.6081 (0.3575)⁎ HHI for MSA −3.3661 (2.1034)

−1.808 (0.4722)⁎⁎⁎ −1.2125 (0.5363)⁎⁎ −0.7171 (0.4564) −2.5544 (2.4623)

−1.7434 (0.4974)⁎⁎⁎ −0.9574 (0.5457)⁎ −0.6634 (0.4951) −2.2635 (2.5602)

−2.315 (0.5482)⁎⁎⁎ −1.852 (0.6561)⁎⁎⁎ −0.4825 (0.4800) −2.5265 (2.6127)

0.0947 (0.1726) −0.0158 (0.0860) −0.0703 (0.0863) −0.5453 (1.0096) 510

0.352 (0.3000) −0.0909 (0.1215) −0.0474 (0.1072) −0.2876 (1.4078) 368

−0.2716 (0.2414) −0.0733 (0.1237) −0.0539 (0.1055) 0.5854 (1.3182) 368

Linear probability model Log population Hospitals per 1000 square mile City pop share of MSA HHI for MSA

Fixed effects model Log population Hospitals per 1000 square mile City pop share of MSA HHI for MSA Number of observations

0.0835 (0.1538) −0.0169 (0.0806) −0.0699 (0.0818) −0.5072 (0.9348) 600

Marginal effects are reported with standard errors in parenthesis. Standard errors are clustered at the city level. Other variables include MSA unionization rate, murder rate, aggravated assault rate, and median household income, level of first response service (ALS/BLS), city population share of MSA, and Herfindahl Index for MSA. ⁎⁎⁎Indicates significance at the 0.001 level, ⁎⁎significance at 0.01, ⁎significance at 0.05.

the results from the city fixed effects specifications are not significant. This regression identifies the parameters of interest from changes over time within city. Population, number of hospitals and land varies more across cities as opposed to within a city across time. In addition, under the fixed effects model, identification relies heavily on cities that experienced massive changes in population; however, such cities are likely to either enter or exit the sample of the 200 largest cities, not allowing for within-city variation in population to affect the choice of EMS configuration. To evaluate the effect of changes in hospital access (hospitals per 1000 square mile) on the city's choice of EMS configuration we use the differences-in-differences estimator, defining a city as being in the ‘treatment’ group if it experienced a below median decline in access. The idea is that while changes to the hospital market (e.g. closure, mergers, elimination of emergency rooms, changes in trauma level designation, etc…) are not driven by EMS provider type, it does affect EMS performance and social costs. As we highlight in our model, cities with worsening hospital infrastructure will be more likely to choose a public–private combination. We find that between 1991 and 1998, regardless of changes in access to hospitals, the prevalence of the public–private model increased in cities. Yet, cities privatized their EMS transport services more frequently when access to their hospital system worsened. Between 1998 and 2003 the prevalence of privatization among cities experiencing diminished access increased, while for cities experiencing improved access, privatization declined. While these results are robust to the inclusion of the observables and statistical models used in previous tables, we decided to report these result in Appendix A as none of the differencesin-differences coefficients are statistically significant.

5. Conclusions In this paper, we have identified dimensions along which combinations of public and private providers of EMS differ and modeled the way a city's characteristics interact with these differences to determine whether a public–private or a pure public combination is preferable for that city. In our conceptual framework, public and private agencies each possess a unique and socially valuable advantage which is fundamental to their status as a public city agency or a private firm. A private firm has the ability to provide service to multiple cities because, unlike a public agency, it is not beholden to one particular city or government, which allows it access to greater economies of scale.20 Advantages of scale of service include reduced average capital, research, and administrative costs as well as the potential for multiple returns to technological innovations of the service. Such advantages make the optimal level of investment in capital and innovation development higher for private firms. Competing with this however, is the superior access of public agencies to existing local infrastructure and resources. Depending on the type of service, this advantage could reduce the costs of establishing basic operation, improve communication and cooperation between different public services, and even enhance performance, due to more detailed knowledge of the local community. Furthermore, we considered the possibility that the preferred provider type may be endogenous to the local characteristics of each particular city. Unlike other literature which discusses provider heterogeneity within a public service, we do not consider such characteristics to be essentially local “tastes,” political or otherwise, nor do we consider mere minimal thresholds of a city characteristic, below which one type of provider is simply not an option. Rather, our framework allows a city's value of the competing advantages of the public and private providers, and hence the determination of the preferred provider type, to vary smoothly with city parameters. We applied our public versus private theory to the provision of emergency medical services, because of its a priori high cost and benefit of both infrastructure and quality improvements. While in first response EMS there is near uniform provision by city fire departments due to the overwhelming cost savings and operational similarities between responding to fires and EMS calls, transport EMS exhibits the very mixed provider equilibrium that is more sensitive to city specific features. It is in transport service that the relative advantage of private firms becomes appealing due to the increased emphasis placed on technology, equipment and training. Our theory predicts that smaller city population, diminished access to hospital care, and preponderance of other cities in close proximity increases the likelihood that a city will choose a public–private EMS configuration, as these factors all increase the importance or strength or the private firm's advantage, relative to the public (fire-department) provider, in providing a superior level of quality improvements. We tested these predictions using a 1991–1998–2005 panel comprised of EMS provider type, demographic, hospital, crime, and unionization data. The effect of access to acute care hospitals was generally borne out by our discrete choice estimations, and the hypotheses about city population and distribution of population within a city's MSA were even more strongly supported. While we focus on EMS, our theory has relevance to the general theory of public service privatization. If the advantages of private firms indeed arise from their scale, then we should expect larger cities to be less likely to utilize private providers for all public services. Also we 20 For example, Emergystat Inc., one of the two largest private providers of emergency ambulance services in Mississippi during our sample period argues (on their company website) that the main benefit of contracting with the company is that the size of their organization's broad southeastern geographic presence provides a competitive advantage over local and regional providers in most areas, including: labor force mobility, coverage, technology enhances quality, reduced costs, retention and recruiting, etc (http://www. emergystat.com/).

G. David, A.J. Chiang / Int. J. Ind. Organ. 27 (2009) 312–319

should expect privatization to be more common in public services which exist in many places, so that there are sufficient opportunities to replicate service and realize returns to scale. Future work could investigate a range of public services and test this concept as a predictor of privatization. Appendix A To evaluate the effect of changes in hospital access (hospitals per 1000 square mile) on the city's choice of EMS configuration, we use the differences-in-differences estimator, defining a city as being in the ‘treatment’ group if it experienced a below median decline in access. We conduct the analysis separately for the 1991–1998 and the 1998– 2005 periods. Because there are likely to be all sorts of factors that cause the EMS system configuration to differ across locations, this lends itself to a difference-in-difference approach.21 The basic conclusions are summarized in Table A below.

Table A Differences-in-differences calculation on the effect of changes in hospital infrastructure 1991 Access worsened 0.3152 (0.4671) Access improved 0.3261 (0.4713) 0.0109 (0.0692)

1998 0.4130 (0.4951) 0.3913 (0.4907) − 0.0217 (0.0691)

−0.0326 (0.1003)

− 0.0486 (0.0974) ✓

DD coefficient Observables Random effects Probit

2005 0.4331 (0.4971) 0.3333 (0.4804) − 0.0998 (0.0995)

−0.0625 (0.1421)

− 0.0505 (0.1414) ✓

Observables Random effects Probit

−0.0326 (0.0460) ✓

1998 Access worsened 0.4076 (0.4929) Access improved 0.3704 (0.4921) −0.0373 (0.1014) DD coefficient

0.0978 (0.0709) 0.0652 (0.0708)

−0.0466 (0.0463) ✓ ✓

−0.2554 (0.4330)

−0.3673 (0.4437) ✓

✓

✓

−0.4482 (0.5986)

−0.3852 (0.6175) ✓

✓

✓

0.0255 (0.0559) −0.0370 (0.1306)

−0.0625 (0.0595) ✓

−0.0490 (0.0591) ✓ ✓

21 Note that the validity of the differences-in-differences estimator is based on the assumption that the underlying ‘trends’ in the outcome variable is the same for both treatment and control groups.

319

References Brown, T.L., Potoski, M., 2003. Managing contract performance: a transaction costs approach. Journal of Policy Analysis and Management 22, 275–297. Chiang, Arthur, David, Guy, Housman, Michael, 2006. The determinants of urban emergency medical services privatization. Critical Planning, vol. 13. summer. Donahue, John D., 1989. The Privatization Decision: Public Ends, Private Means. Basic Books, New York, NY. Eckstein, Marc, Pratt, Franklin, 2002. Chapter 6: fire, In: Kuehl, Alexander E. (Ed.), Prehospital Systems and Medical Oversight, 3rd Edition. Fernandez, Sergio, 2005. Accounting for Performance in Contracting for Services: Are Successful Contractual Relationships Controlled or Managed? Working paper. Hart, Oliver, Shleifer, Andrei, Vishny, Robert W., 1997. The proper scope of government: theory and an application to prisons. The Quarterly Journal of Economics 112 (4), 1127–1161. Hefetz, Amir, Warner, Mildred E., 2004. Privatization and its reverse: explaining the dynamics of the government contracting process. Journal of Public Administration Research and Theory 14 (2), 171–190. Hirsch, W., 1995. Factors important in local government's privatization decisions. Urban Affairs Review 31, 226–243. Hirsh, Barry T., Macpherson, David A., 2003. Union Membership and Earnings Data Book, 2003 ed. The Bureau of National Affairs. Journal of Emergency Medical Services (JEMS) 1991, 1998, and 2005 Annual 200-City Surveys. Levin, Jonathan, Tadelis, Steven, 2007. Contracting for Government Services: Theory and Evidence from U.S. Cities. working paper, August. Miranda, Rowen, Lerner, Allen, 1995. Bureaucracy, organizational redundancy, and the privatization of public services. Public Administration Review 55 (2), 193–200. Persse, David E., 2002. Chapter 19: response choices, In: Kuehl, Alexander E. (Ed.), Prehospital Systems and Medical Oversight, 3rd Edition. Sclar, E.D., 2000. You Don't Always Get What You Pay For: The Economics of Privatization. Cornell University Press, Ithaca. Warner, Mildred, Hefetz, Amir, 2003. Rural — urban differences in privatization: limits to the competitive state. Government and Policy 21, 703–718. Williamson, Oliver E., 1985. The Economic Institutions of Capitalism. Free Press, New York.